How often do you see fake chips?

[ar__systems]

Funny story.

A customer came back to me with problems they had with a production batch of a design I completed few years ago. Suddenly large proportion of the batch was not working properly.

Years of working in software engineering have taught me to always first to look for my own screw-ups, since that is usually the most common cause for any problems So calls like that always worry me. Did I miss something?

The device is a low power contamination detector that shines a light through a sample and measures the response. Quick tests showed the photodetector side was not working properly. The circuit is a standard transimpedance amplifier straight from datasheet. In absence of photo-current through the diode it works as a unity gain buffer, but on the affected device the output was less that than the non-inverting input. How is that even possible? Transplanting the questionable opamp onto test PCB showed that it is actually working, but the input bias current specs is not met. The behavior was consistent with parasitic current between VDD and inverting input. The amplifier used in the device is LTC6244, that has pA bias currents, but I was observing currents in the range of uA. Having experience with low power circuits in the past, I started to suspect PCB surface leakage. That was quickly ruled out, since VDD and OPA1- are on opposite sides of the package and there are traces separating them. ESD damage? Plausible, but there was more than one board with EXACTLY the same behaviour. Replacing the funny amplifier with a new one fixed all of them. When the number of boards showing the same strange behavior became 10 out 10, I had to rule out ESD as well. If the pin is not just shorted to VDD or VSS, what are the chances of developing identical leakage on 10 boards by ESD?

Chips looked ok, marking looked legit. But, I saw that lead frame protrusions were in different locations compared to the known proper chip. But ok, my proper chips were purchased 3 years ago. Maybe the company has made a chip revision or something? And then I cranked up zoom on my microscope to the maximum. Whoa! I've never seen that on any other chips, ever! And it was not just one of the like that. They were all like that! Bond wires sticking out? :facepalm:

Interestingly, my client is not cheap, they are doing all of the assembly locally in Canada. Who would've thunk that the assembly house would source components on AliExpress

[wraper]

Bond wires sticking out? :facepalm:

They should be some other chip with top surface sanded and new marking applied with laser. As this package is rather thin to begin with, it's not that surprising they almost reached bond wires, and then laser completely exposed them.

[ar__systems]

Bond wires sticking out? :facepalm:

They should be some other chip with top surface sanded and new marking applied with laser. As this package is rather thin to begin with, it's not that surprising they almost reached bond wires, and then laser completely exposed them.

Makes sense! I did not think of that. But yes, these are clearly lower performance opamps made looked like pretty advanced LTC6244.

[DerekG]

They were all like that! Bond wires sticking out?

Yes, this is typical of chips that have their tops ground down.

There is another thread running re fake chips:

https://www.eevblog.com/forum/eda/knock-off-chips/msg2274432/#msg2274432

[Psi]

You will eventually come across fake chip if you use low price/common hobby assemblers.  eg like PCBway, elecrow , etc..

Fake parts are common when the following are true.
- The chip is quite expensive variant
- It's a category of chip that is common and usually cheap.

An example is an expensive automotive rated 3.3 or 5V voltage regulator.

It's easy for someone out there to re-stamp a batch of generic cheap 5V regulator with the expensive part number and sell them for a quick profit.
This sort of fake is not picked up very quickly because it does works fine as a 5V reg.
You might end up with a product missing the reverse polarity protection or missing the built in 60V transient protection you expected in the automotive reg.


Ways to prevent this happening to you.
- If possible, pick a part with a less common footprint (This makes your job harder if you need to change it later, but much less likely to get fakes).
The people faking chips will pick a common footprint for their fakes. Easy for them to sell that way.

- If there is a critical part, make a point of telling the assembler that the part MUST be purchased from Digikey/Mouser etc.. and that you require them to provide the invoice for it.

- For critical parts purchase them yourself and send to the assembler.

- Place critical part yourself if low volume and feasible.

- For critical parts make sure you test your finished boards in ways that would uncover if parts don't have the features you expect. eg reverse polarity test if your Vreg is supposed to have it.   You can get boards tested by the PCBA house.

[ar__systems]

- If there is a crucial part, make a point of telling the assembler that the part MUST be purchased from Digikey/Mouser etc.. and that you require them to provide the invoice for it.

Yep, I'm going to tell all of my customers to always insist on seeing invoices for the purchased parts. In case of this customer there was another suspect component on the board (LED), and I told them to get the invoice for it. Apparently the assembly house was not able to produce it 

[Bud]

Interestingly, my client is not cheap, they are doing all of the assembly locally in Canada. Who would've thunk that the assembly house would source components on AliExpress

Some of those may simply be a order handling office with assembly done in China. I' ve seen such ones at local tech shows.

[OwO]

- For critical parts make sure you test your finished boards in ways that would uncover if parts don't have the features you expect. eg reverse polarity test if your Vreg is supposed to have it.   You can get boards tested by the PCBA house.

I think this is the best advice, often sourcing from low volume distributors (digikey etc) for production is way too expensive and you have no choice but to use alternative supply chains. Testing is the only way to be sure. Personally I have not seen fake chips, I do use high speed op-amps from shady sources but test for key specs in circuit.

[thm_w]

The font looks slightly bolder than the image from here: https://octopart.com/ltc6244hms8%23pbf-analog+devices-88320134
But it would be hard to say that is justification.

Bond wires thing is hilarious though..

[Psi]

You also do have to accept that sometimes you WILL get fake chips and they will work totally fine and you will never actually know.

[Smallsmt]

Some years before we bought some ONSEMI DC/DC converter chips.
They looked exactly like the originals only missing the tolerances so converter needed more current than our 15000 we bought before.
We tracked the LOT number down to manufacturer and they told us maybe somebody sold scrap goods they sorted out.

Really crazy situation too.

[BreakingOhmsLaw]

It's not just expensive stuff, too.  I've seen a case where a ten cent through-hole FET was a relabelled PNP tranny.

[Stray Electron]

 I've been looking for a particular out of production RF prescaler IC and they're readily available on E-bay and some of the other common sources but EVERYONE of them that I've checked has been counterfeit.  I've tried sources in the US, Canada, China, Hong Kong and Germany and none have original production parts.  The problem of fake and counterfeit parts is world wide. It's so bad that I've given up buying parts unless they're still in production and available from a reputable dealer like Digikey. For old parts, I hunt down an old piece of equipment that used them and take the parts from there.

[Synthtech]

In the electronic music niche there are fake chips all over EBay. Bucket brigade chips, fake CA3280’S, fake SSM & Curtis chips, fake power transistors and so on.

[m3vuv]

seen lots of fake 2n3055 tranys from ebay that short at 1 amp.

[CatalinaWOW]

I've tried sources in the US, Canada, China, Hong Kong and Germany and none have original production parts.

If someone sells a counterfeit, they designed it to work, or at least barely works. If one doesn't work, it's a fluke, and the other one should work.
If throughout all your sources all counterfeit chips in your circuit don't work, your circuit is designed poorly that it depends on boundary performance or even out of spec performance to work properly.

This is true for the high end of the counterfeit market.  High defined by quality, not necessarily price.  But that is only part of the counterfeit market.  Other parts of the market include selling the design that was intended to work, but didn't quite make it over the bar, selling the design that wasn't fully understood or was 'improved' by someones definition of better, remarking lower grade parts (selling B or C parts as A), packaging culls (parts that failed production test, but may partly operate), and just flat out lying about the content (Remarking anything with the right package type.)     

One might argue that requiring tightly tested parts (A parts) is poor circuit design, but this argument does not apply to the worse forms of counterfeits.  For old or rare parts which may have changed hands many times before reaching the final user there is no complete defense  against counterfeit parts other than thorough testing.  Your situation will tell whether component level or circuit level testing is better.

[L_Euler]

I ordered a lot of 10 CA3080 transcon amps from a Chinese supplier.  10 out of 10 bad.  Don't know if they were counterfeits or fails, but thankfully they were cheap and the loss was minimal.  I got what I paid for.

[Bud]

In the electronic music niche there are fake chips all over EBay. Bucket brigade chips, fake CA3280’S, fake SSM & Curtis chips, fake power transistors and so on.

I do not think anyone make bucket brigade chips anymore, it is a lost art like nixie tubes. Those clones on the market are likely utilize some sort of emulation.

[Alex Eisenhut]

Yeah, terrible.

[thm_w]

Left: NE5532 from aliexpress seller Vanxy. $1 for 20pc, too good to be true price, so expected fakes, but I'm disappointed no other buyers had tested them properly (142 orders and no bad reviews!). I've seen good knowledgeable reviews on some parts, maybe opamps are generic enough to attract a different crowd. But they are a reasonable seller otherwise, and quickly refunded. Has dimple on pin 1. Slew rate ~2.4V/us and works on a single supply.
Right: NE5532 from LCSC. $3.80 for 20pc. No dimple. Silkscreen is similar enough, looks worse quality if anything. Does not work properly on single supply, as expected.

I'm making the assumption the LCSC part is legit, I haven't done proper testing on it.
Digikey is about $20 for 20pc, or even TI direct is $12.

Will try taobao next.

[james_s]

If you get verified fake/counterfeit parts, always apply for a refund. If enough people do this it will not be profitable to keep selling fake crap. The counterfeiters rely on the fact that most people will just write off a small purchase.

[james_s]

You can open a claim against the seller for selling counterfeit goods, then ebay will issue you the refund and ding the seller. Enough claims and they get kicked off.

[T3sl4co1l]

That part was clearly desoldered (salvaged).

(Unless you took the picture after testing it, in which case that's technically correct but it's your own doing, so nevermind?...)

The white bar [not] over the dimple is amusing, surely no real manufacturer does that (I don't know of any examples of this?).

LM358?

Tim

[OwO]

And if you do the refund, you are no longer allowed to leave a feedback, that's why eBay fake sellers can keep 100% positive reviews.

Maybe that's why I had a bad experience with ebay but much better with aliexpress. As long as you stick to rating > 98% on aliexpress usually you don't get duds, but op amps are just one of the things you shouldn't buy from the marketplace and expect it to be genuine. Actually I don't expect anything I buy to be genuine, I just expect them to mostly perform within spec and it turns out to be true for most parts. Linear regulators and op amps are the exception.

[thm_w]

And if you do the refund, you are no longer allowed to leave a feedback, that's why eBay fake sellers can keep 100% positive reviews.

Didn't know that about ebay. aliexpress lets you get a refund and leave a negative review, thankfully.

That part was clearly desoldered (salvaged).
(Unless you took the picture after testing it, in which case that's technically correct but it's your own doing, so nevermind?...)
The white bar [not] over the dimple is amusing, surely no real manufacturer does that (I don't know of any examples of this?).

LM358?

Tim

Sorry should have specified this was after I desoldered it from the test board. You are probably right that its a LM358 or something close.
Although I'm going to try my luck with a few salvaged parts because: it seems faking a used part would be a poor choice as you'd get more money for one sold as "new", and salvaged parts are sometimes ridiculously cheap say 10x less cost than new. Even if 8 out of 10 work I would be happy, as long as its not hard to validate their performance.
We recycle a lot of electronics and 99% of the ICs on the board are working well.

Maybe that's why I had a bad experience with ebay but much better with aliexpress. As long as you stick to rating > 98% on aliexpress usually you don't get duds, but op amps are just one of the things you shouldn't buy from the marketplace and expect it to be genuine. Actually I don't expect anything I buy to be genuine, I just expect them to mostly perform within spec and it turns out to be true for most parts. Linear regulators and op amps are the exception.

Agreed.
Any opamps/regulators I buy will be from LCSC/digikey unless its some unique package/pinout/function.